Grok Bot engineer's 45-minute lecture: inside a background personal agent
A SpaceX engineer walks through Grok Bot's internals: a second model checking every command, the profile and work log it keeps, sleeping between tasks, its own cloud Linux box, and phishing and password handling.
Anthropic, OpenAI and Meta now all have AI agents that keep working after you close your laptop. SpaceX has one too, and an engineer who works on Grok Bot shows what's inside it in a 45-min lecture. 04:57 - a second AI model that checks every command 08:29 - the profile and work log it keeps on you 10:13 - why the agent sleeps between tasks 12:19 - its own Linux computer in the cloud 26:29 - phishing emails, approvals and passwords Paid AI agent courses rarely show this much of a real product, and this lecture is free. Watch it first, then read the article below and set up Claude Duet in 10 steps.
This post quotes another post: x.com/0xkense1/status/2113020432391520378